Figure legends
Figure 1 : Deviations of monthly mean temperature and precipitation for 2016 (blue bars, panel A) and 2017 (red bars, panel B) calculated as the difference from the 2000-2015 average value at the site. Temperature of April 25th, 2016 measured at canopy level (24 m) is reported in inset graph of panel a , while the annual precipitation of the year 2016 (blue dots), 2017 (red dots) and the long-term average (black dots) is reported in the inset graph of panel b .
Figure 2 : Seasonal dynamics of Leaf area index (LAI, m2 m–2, panel a ) and daily stem radial increment (panel b) for the years 2016, 2017 and the 2000-2015 reference period. LAI was derived from Moderate Resolution Imaging Spectroradiometer (MODIS, see Materials and Methods), for 2016 (blue line), 2017 (red line) and for the 2000-2015 reference period (black line). Solid lines are the modelled LAI pattern, using two logistic functions for the increasing and decreasing phases. Dots are the raw MODIS-LAI values. In panel b the daily radial increment for 2016 (blue dots) and 2017 (red dots) are shown, while the inset graph reports the long-term series data of Basal Area chronology (BAI, cm2 year–1), where the last two dots represent the BAI value obtained in 2016 (blue dot) and 2017 (red dot), respectively.
Figure 3 : Phenological data for the experimental beech forest site (top panel) and seasonal dynamic of NSCs content as total NSCs (panel a ), starch (panel b ) and soluble sugars (panelc ).In the top panel, the different colours represent dormancy (dark orange), the period between the green up and the maximum Leaf area index (LAI, m2 m–2) value (light green), the maximum LAI (dark green), the senescence phase (light orange) and, finally, the leafless period after the late frost in 2016 (grey). In the panels a, b, and c, blue and red dots represent carbohydrate concentrations of 2016 and 2017, respectively, while the black lines and grey area show modelled intra annual dynamic of carbohydrates and 95 % interval of confidence (1.96 SE), respectively. Each point is the mean of five beech trees and bars are the intervals of confidence (1.96 SE) (see Material and Methods). Modelled values are derived from 39 and 28 measurements of starch and soluble sugar content, respectively.